Label generation method for online learning channel sounding by using pseudo-random data

Abstract

A wireless transmit / receive unit (WTRU) may receive pseudo-random data (PRD), one or more channel state information reference signals (CSI-RS), and / or one or more learning reference signals (L-RS). The WTRU may determine a CSI measurement result based on the received CSI-RS and / or determine an L-RS measurement result based on the received L-RS. The WTRU may generate one or more CSI tags based on the PRD, the CSI-RS measurements, and / or the L-RS measurements. The WTRU may generate one or more output CSI using an artificial intelligence / machine learning (AI / ML) model and based on the PRD and / or CSI-RS measurements. The WTRU may train an AI / ML model using the one or more CSI tags and the one or more output CSI.

Description

[0001] Cross-references to related applications This application claims the benefit of U.S. Provisional Patent Application No. 63 / 584,204, filed September 21, 2023, the entire contents of which are incorporated herein by reference. Background Technology

[0002] Artificial intelligence (AI) can be broadly defined as the behavior exhibited by machines that mimic cognitive functions to sense, reason, adapt, and / or act. AI components can involve learning through data without explicitly configuring sequences of action steps to achieve behavior and / or fulfill requirements. Such AI components can enable the learning of complex behaviors that might be difficult to specify and / or achieve using traditional methods.

[0003] Machine learning (ML) can refer to a type of algorithm that solves problems based on learning from experience (e.g., data) without explicit programming (e.g., configuring a set of rules). ML can be considered a subset of AI. Different ML paradigms can be envisioned based on the nature of the data available to the learning algorithm and / or the feedback. In one example, supervised learning methods may involve learning a function that maps inputs to outputs based on labeled training examples. Each training example can be, for example, a pair consisting of an input and its corresponding output. In another example, unsupervised learning methods may involve detecting patterns in data without pre-existing labels. In yet another example, reinforcement learning methods may involve performing a sequence of actions in an environment to maximize cumulative reward. Summary of the Invention

[0004] A Wireless Transmit / Receive Unit (WTRU) can receive pseudo-random data (PRD), one or more Channel State Information Reference Signals (CSI-RS), and / or one or more Learning Reference Signals (L-RS). The WTRU can determine CSI measurement results based on the received CSI-RS measurements. The WTRU can determine L-RS measurement results based on the received L-RS. The WTRU can generate one or more CSI tags based on the PRD, CSI-RS measurement results, and / or L-RS measurement results. The WTRU can use an Artificial Intelligence / Machine Learning (AI / ML) model and generate one or more output CSIs based on the PRD and / or CSI-RS measurement results. The WTRU can train an AI / ML model using one or more CSI tags and one or more output CSIs.

[0005] WTRU can determine the variation of the loss function of an AI / ML model based on one or more CSI labels and one or more output CSIs. WTRU can determine whether training is complete based on a comparison of the loss function value with a threshold. The loss function of an AI / ML model can determine a measure of training accuracy during the training of the AI / ML model.

[0006] WTRU can determine a model drift metric. WTRU can determine the performance of an AI / ML model used for CSI estimation based on model drift by applying statistical tests to CSI-RS measurements. WTRU can retrain the AI / ML model if the model drift metric exceeds a drift threshold.

[0007] WTRU can send a report to the network in response to an indication that a model drift metric exceeds the threshold. This report may include details of the model drift or recommendations for adjusting the AI / ML model.

[0008] The WTRU can send reports to the network based on CSI-RS measurement results and / or requests to be configured with one or more training assignments. The WTRU can receive from the network an indication (e.g., a response) regarding whether the request to be configured with one or more training assignments is permitted. This indication (e.g., the response) may also include seed information for generating the PRD.

[0009] The WTRU can receive PRD training allocation configurations. PRD training allocations can indicate a first type, a second type, and / or a third type of PRD training allocation. A first type of PRD training allocation configures the WTRU to use PRD and / or CSI-RS measurements for a first set of resource blocks carrying PRDs and CSI-RS. A second type of PRD training allocation configures the WTRU to use PRD and / or CSI-RS measurements for a second set of resource blocks containing data carrying PRDs, CSI-RS, and / or not carrying PRDs. PRDs and data-including resource elements without PRDs can occupy the same resource blocks. A third type of PRD training allocation configures the WTRU to use PRD and / or CSI-RS measurements for a third set of resource blocks containing data carrying PRDs, CSI-RS, and / or not carrying PRDs. PRDs and data-including resource elements without PRDs can occupy different but overlapping resource blocks.

[0010] The allocation of PRD resources of the first, second and / or third types can be signaled in the downlink control information (DCI), semi-statically via system information blocks and / or via broadcast messages.

[0011] The WTRU can send a first message including an online training request. The online training request may indicate the number of resource blocks carrying the PRD for training, channel statistics measurements, channel diversification requests, and / or the priority of the requests. The WTRU can receive a second message including an online training response. The online training response may indicate the location of the resource blocks including the PRD, the allocated online training duration, the allocated online training start time, and / or the allocated online training learning reference signal (L-RS). Attached Figure Description

[0012] Figure 1A This is a system diagram illustrating an example communication system in which one or more of the disclosed embodiments can be implemented.

[0013] Figure 1B The illustration is based on an embodiment and can be used for Figure 1A The diagram shows a system diagram of an example wireless transmit / receive unit (WTRU) within a communication system.

[0014] Figure 1C The illustration is based on an embodiment and can be used for Figure 1A The diagram shows an example radio access network (RAN) and an example core network (CN) within the communication system.

[0015] Figure 1D The illustration is applicable to embodiments. Figure 1A The system diagram shows another example RAN and another example CN within the communication system shown.

[0016] Figure 2 This is an example of Type 1 training allocation using pseudo-random data (PRD) and channel state information reference symbols (CSI-RS).

[0017] Figure 3 This is an example of Type 2 training allocation utilizing PRD, CSI-RS, and / or synchronous data scheduling.

[0018] Figure 4 This is an example graph of online data labeling and / or training for a CSI estimator based on artificial intelligence / machine learning (AI / ML).

[0019] Figure 5 This is an example graph showing the performance monitoring of an AI / ML-based CSI estimator using PRD.

[0020] Figure 6 This is an example graph showing the performance monitoring of an AI / ML-based CSI estimator using a learned reference signal (L-RS).

[0021] Figure 7An example AI / ML CSI estimation module is described from the perspective of WTRU. Detailed Implementation

[0022] Figure 1A This diagram illustrates an example communication system 100 in which one or more of the disclosed embodiments may be implemented. The communication system 100 may be a multiple access system that provides content, such as voice, data, video, messaging, broadcasting, etc., to multiple wireless users. The communication system 100 enables multiple wireless users to access this content by sharing system resources (including wireless bandwidth). For example, the communication system 100 may employ one or more channel access methods, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal FDMA (OFDMA), Single Carrier FDMA (SC-FDMA), Zero-Tail Unique Word DFT Extended OFDM (ZT UW DTS-s OFDM), Unique Word OFDM (UW-OFDM), Resource Block Filtered OFDM (OFDM), Filter Bank Multicarrier (FBMC), etc.

[0023] like Figure 1A As shown, the communication system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, RAN 104 / 113, CN 106 / 115, Public Switched Telephone Network (PSTN) 108, Internet 110, and other networks 112. However, it should be understood that the disclosed embodiments contemplate any number of WTRUs, base stations, networks, and / or network elements. Each of WTRUs 102a, 102b, 102c, and 102d may be any type of device configured to operate and / or communicate in a wireless environment. For example, WTRUs 102a, 102b, 102c, and 102d (any of which may be referred to as a “station” and / or “STA”) may be configured to transmit and / or receive wireless signals and may include user equipment (UE), mobile stations, fixed or mobile subscriber units, subscription-based units, pagers, cellular phones, personal digital assistants (PDAs), smartphones, laptops, netbooks, personal computers, wireless sensors, hotspots or Mi-Fi devices, Internet of Things (IoT) devices, watches or other wearable devices, head-mounted displays (HMDs), vehicles, drones, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in industrial and / or automated processing chain environments), consumer electronics devices, devices operating on commercial and / or industrial wireless networks, etc. Any of WTRUs 102a, 102b, 102c, and 102d may be interchangeably referred to as WTRUs.

[0024] The communication system 100 may also include base station 114a and / or base station 114b. Each of base stations 114a and 114b can be any type of device configured to wirelessly interface with at least one of WTRUs 102a, 102b, 102c, and 102d to facilitate access to one or more communication networks, such as CN 106 / 115, the Internet 110, and / or other networks 112. For example, base stations 114a and 114b can be base transceiver stations (BTS), Node-B, eNode B, home Node B, home eNode B, gNB, NR Node B, site controllers, access points (APs), wireless routers, etc. Although base stations 114a and 114b are depicted as single elements, it should be understood that base stations 114a and 114b can include any number of interconnected base stations and / or network elements.

[0025] Base station 114a may be part of RAN 104 / 113, which may also include other base stations and / or network elements (not shown in the figures), such as base station controllers (BSCs), radio network controllers (RNCs), relay nodes, etc. Base station 114a and / or base station 114b may be configured to transmit and / or receive radio signals on one or more carrier frequencies, which may be referred to as cells (not shown in the figures). These frequencies may be located in licensed spectrum, unlicensed spectrum, or a combination of licensed and unlicensed spectrum. A cell may provide radio service coverage for a specific geographic area, which may be relatively fixed or may vary over time. A cell may be further divided into cell sectors. For example, the cell associated with base station 114a may be divided into three sectors. Therefore, in one embodiment, base station 114a may include three transceivers, one for each sector of the cell. In embodiments, base station 114a may employ multiple-input multiple-output (MIMO) technology and may use multiple transceivers for each sector of the cell. For example, beamforming may be used to transmit and / or receive signals in desired spatial directions.

[0026] Base stations 114a and 114b can communicate with one or more of WTRUs 102a, 102b, 102c, and 102d via air interface 116, which can be any suitable wireless communication link (e.g., radio frequency (RF), microwave, centimeter wave, micrometer wave, infrared (IR), ultraviolet (UV), visible light, etc.). Air interface 116 can be established using any suitable radio access technology (RAT).

[0027] More specifically, as described above, the communication system 100 can be a multiple access system and can employ one or more channel access schemes, such as CDMA, TDMA, FDMA, OFDMA, SC-FDMA, etc. For example, base stations 114a and WTRUs 102a, 102b, and 102c in RAN 104 / 113 can implement radio technologies such as Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access (UTRA), which can use Wideband CDMA (WCDMA) to establish air interfaces 115 / 116 / 117. WCDMA can include communication protocols such as High-Speed ​​Packet Access (HSPA) and / or Evolved HSPA (HSPA+). HSPA can include High-Speed ​​Downlink (DL) Packet Access (HSDPA) and / or High-Speed ​​UL Packet Access (HSUPA).

[0028] In the embodiment, base station 114a and WTRUs 102a, 102b, 102c can implement radio technologies such as Evolved UMTS Terrestrial Radio Access (E-UTRA), which can use Long Term Evolution (LTE) and / or LTE-Advanced (LTE-A) and / or LTE-Advanced Pro (LTE-A Pro) to establish air interface 116.

[0029] In the embodiment, base station 114a and WTRUs 102a, 102b, 102c can implement radio technologies such as NR radio access, which can use new radio (NR) to establish air interface 116.

[0030] In the embodiments, base station 114a and WTRUs 102a, 102b, and 102c can implement multiple radio access technologies. For example, base station 114a and WTRUs 102a, 102b, and 102c can simultaneously implement LTE radio access and NR radio access, for example, using the dual connectivity (DC) principle. Therefore, the air interface utilized by WTRUs 102a, 102b, and 102c can be characterized by multiple types of radio access technologies and / or transmissions to and from multiple types of base stations (e.g., eNBs and gNBs).

[0031] In other embodiments, base station 114a and WTRUs 102a, 102b, 102c can implement radio technologies such as IEEE 802.11 (i.e., Wi-Fi), IEEE 802.16 (i.e., WiMAX), CDMA2000, CDMA2000 1X, CDMA2000 EV-DO, Provisional Standard 2000 (IS-2000), Provisional Standard 95 (IS-95), Provisional Standard 856 (IS-856), Global System for Mobile Communications (GSM), Enhanced Data Rate GSM Evolution (EDGE), and GSM EDGE (GERAN).

[0032] Figure 1A Base station 114b can be, for example, a wireless router, a home Node B, a home eNode B, or an access point, and can utilize any suitable RAT to facilitate wireless connectivity in local areas such as commercial locations, homes, vehicles, campuses, industrial facilities, air corridors (e.g., for drone use), roads, etc. In one embodiment, base station 114b and WTRUs 102c, 102d can implement radio technologies such as IEEE 802.11 to establish a wireless local area network (WLAN). In another embodiment, base station 114b and WTRUs 102c, 102d can implement radio technologies such as IEEE 802.15 to establish a wireless personal area network (WPAN). In yet another embodiment, base station 114b and WTRUs 102c, 102d can utilize cellular-based RATs (e.g., WCDMA, CDMA2000, GSM, LTE, LTE-A, LTE-A Pro, NR, etc.) to establish picocells or femtocells. Figure 1A As shown, base station 114b can be directly connected to Internet 110. Therefore, base station 114b can access Internet 110 without going through CN 106 / 115.

[0033] RAN 104 / 113 can communicate with CN 106 / 115, which can be any type of network configured to perform the following operations: provide voice, data, application, and / or Voice over Internet Protocol (VoIP) services to one or more of WTRUs 102a, 102b, 102c, and 102d. The data may have different Quality of Service (QoS) requirements, such as different throughput requirements, latency requirements, fault tolerance requirements, reliability requirements, data throughput requirements, mobility requirements, etc. CN 106 / 115 can provide call control, billing services, location-based services, prepaid calling, internet connectivity, video distribution, etc., and / or perform advanced security functions such as user authentication. Although Figure 1A Although not shown in the diagram, it should be understood that RAN104 / 113 and / or CN106 / 115 can communicate directly or indirectly with other RANs that use the same or different RATs as RAN 104 / 113. For example, in addition to connecting to RAN 104 / 113, which may utilize NR radio technology, CN 106 / 115 can also communicate with another RAN (not shown in the diagram) that uses GSM, UMTS, CDMA 2000, WiMAX, E-UTRA, or WiFi radio technology.

[0034] CN 106 / 115 can also act as a gateway for WTRU 102a, 102b, 102c, 102d to access PSTN 108, the Internet 110, and / or other networks 112. PSTN 108 may include a circuit-switched telephone network providing Common Old-Style Telephone Service (POTS). The Internet 110 may include a global system of interconnected computer networks and devices using common communication protocols such as Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and / or Internet Protocol (IP) from the TCP / IP Internet Protocol suite. Network 112 may include wired and / or wireless communication networks owned and / or operated by other service providers. For example, network 112 may include another CN connected to one or more RANs, which may employ the same or different RAT as RAN 104 / 113.

[0035] Some or all of the WTRUs 102a, 102b, 102c, and 102d in the communication system 100 may include multi-mode capability (e.g., WTRUs 102a, 102b, 102c, and 102d may include multiple transceivers for communicating with different wireless networks via different wireless links). For example, Figure 1AThe WTRU 102c shown can be configured to communicate with a base station 114a that may employ cellular-based radio technology and a base station 114b that may employ IEEE 802 radio technology.

[0036] Figure 1B This is a system diagram illustrating the example WTRU 102. (Example: ...) Figure 1B As shown, WTRU 102 may include a processor 118, a transceiver 120, a transmitting / receiving element 122, a speaker / microphone 124, a keyboard 126, a display / touchpad 128, non-removable memory 130, removable memory 132, a power supply 134, a Global Positioning System (GPS) chipset 136, and / or other peripheral devices 138, etc. It should be understood that, while remaining consistent with the embodiments, WTRU 102 may include any sub-combination of the above-described elements.

[0037] Processor 118 can be a general-purpose processor, a special-purpose processor, a conventional processor, a digital signal processor (DSP), multiple microprocessors, one or more microprocessors associated with a DSP core, a controller, a microcontroller, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) circuit, any other type of integrated circuit (IC), a state machine, etc. Processor 118 can perform signal encoding / decoding, data processing, power control, input / output processing, and / or any other function that enables WTRU 102 to operate in a wireless environment. Processor 118 can be coupled to transceiver 120, and transceiver 120 can be coupled to transmitting / receiving element 122. Although Figure 1B While the processor 118 and transceiver 120 are depicted as separate components, it should be understood that the processor 118 and transceiver 120 may be integrated together in an electronic package or chip.

[0038] Transmitting / receiving element 122 may be configured to transmit signals to or receive signals from a base station (e.g., base station 114a) via air interface 116. For example, in one embodiment, transmitting / receiving element 122 may be an antenna configured to transmit and / or receive RF signals. In another embodiment, transmitting / receiving element 122 may be a transmitter / detector configured to transmit and / or receive, for example, IR, UV, or visible light signals. In yet another embodiment, transmitting / receiving element 122 may be configured to transmit and / or receive both radio frequency signals and optical signals. It should be understood that transmitting / receiving element 122 may be configured to transmit and / or receive any combination of wireless signals.

[0039] Although Figure 1BWhile the transmitting / receiving element 122 is depicted as a single element, the WTRU 102 may include any number of transmitting / receiving elements 122. More specifically, the WTRU 102 may employ MIMO technology. Thus, in one embodiment, the WTRU 102 may include two or more transmitting / receiving elements 122 (e.g., multiple antennas) for transmitting and receiving wireless signals via the air interface 116.

[0040] Transceiver 120 can be configured to modulate the signal to be transmitted by transmitting / receiving element 122 and demodulate the signal received by transmitting / receiving element 122. As described above, WTRU 102 can have multimode capability. Therefore, transceiver 120 can include multiple transceivers to enable WTRU 102 to communicate via various RATs (such as, for example, NR and IEEE 802.11).

[0041] The processor 118 of WTRU 102 can be coupled to a speaker / microphone 124, a keyboard 126, and / or a display / touchpad 128 (e.g., a liquid crystal display (LCD) unit or an organic light-emitting diode (OLED) display unit) and can receive user input data from these devices. The processor 118 can also output user data to the speaker / microphone 124, keyboard 126, and / or display / touchpad 128. Furthermore, the processor 118 can access information and store data from any type of suitable memory, such as non-removable memory 130 and / or removable memory 132. Non-removable memory 130 may include random access memory (RAM), read-only memory (ROM), a hard disk, or any other type of memory storage device. Removable memory 132 may include a subscriber identity module (SIM) card, memory stick, secure digital storage (SD) card, etc. In other embodiments, the processor 118 can access information and store data from memory that is not physically located on WTRU 102 (such as on a server or home computer (not shown)).

[0042] The processor 118 can receive power from the power supply 134 and can be configured to distribute power to other components in the WTRU 102 and / or control the power going to other components. The power supply 134 can be any device suitable for powering the WTRU 102. For example, the power supply 134 may include one or more dry cell batteries (e.g., nickel-cadmium (NiCd), nickel-zinc (NiZn), nickel-metal hydride (NiMH), lithium-ion (Li-ion), etc.), solar cells, fuel cells, etc.

[0043] The processor 118 may also be coupled to a GPS chipset 136, which may be configured to provide location information (e.g., longitude and latitude) regarding the current location of the WTRU 102. Attached to or replacing the information from the GPS chipset 136, the WTRU 102 may receive location information from base stations (e.g., base stations 114a, 114b) via air interface 116, and / or determine its location based on the timing of signals received from two or more nearby base stations. It should be understood that, while remaining consistent with the embodiments, the WTRU 102 may obtain location information using any suitable location determination method.

[0044] The processor 118 may also be coupled to other peripheral devices 138, which may include one or more software and / or hardware modules that provide additional features, functionality, and / or wired or wireless connectivity. For example, peripheral devices 138 may include accelerometers, electronic compasses, satellite transceivers, digital cameras (for photos and / or video), Universal Serial Bus (USB) ports, vibration devices, television transceivers, hands-free headsets, Bluetooth® modules, FM radio units, digital music players, media players, video game player modules, internet browsers, virtual reality and / or augmented reality (VR / AR) devices, activity trackers, etc. Peripheral devices 138 may include one or more sensors, which may be one or more of the following: gyroscopes, accelerometers, Hall effect sensors, magnetometers, orientation sensors, proximity sensors, temperature sensors, time sensors, geolocation sensors; altimeters, light sensors, touch sensors, magnetometers, barometers, gesture sensors, biosensors, and / or humidity sensors.

[0045] WTRU 102 may include a full-duplex radio, for which the transmission and reception of some or all signals (e.g., signals associated with specific subframes for both UL (e.g., for transmission) and downlink (e.g., for reception)) may be concurrent and / or simultaneous. The full-duplex radio may include an interference management unit 139 for reducing and / or substantially eliminating self-interference via hardware (e.g., a choke) or through signal processing by a processor (e.g., a separate processor (not shown) or through processor 118). In embodiments, WTRU 102 may include a half-duplex radio, for which the transmission and reception of some or all signals (e.g., signals associated with specific subframes for UL (e.g., for transmission) or downlink (e.g., for reception)) may be concurrent and / or simultaneous.

[0046] Figure 1CThis is a system diagram illustrating RAN 104 and CN 106 according to an embodiment. As described above, RAN 104 can communicate with WTRUs 102a, 102b, and 102c via air interface 116 using E-UTRA radio technology. RAN 104 can also communicate with CN 106.

[0047] RAN 104 may include eNodeBs 160a, 160b, and 160c, but it should be understood that RAN 104 may include any number of eNodeBs while remaining consistent with the embodiments. eNodeBs 160a, 160b, and 160c may each include one or more transceivers for communicating with WTRUs 102a, 102b, and 102c via air interface 116. In one embodiment, eNodeBs 160a, 160b, and 160c may implement MIMO technology. Therefore, for example, eNodeB 160a may use multiple antennas to transmit radio signals to and / or receive radio signals from WTRU 102a.

[0048] Each of eNodeBs 160a, 160b, and 160c can be associated with a specific cell (not shown in the diagram) and can be configured to handle radio resource management decisions, handover decisions, and user scheduling in the UL and / or DL, etc. Figure 1C As shown, eNodeB 160a, 160b, and 160c can communicate with each other via the X2 interface.

[0049] Figure 1C The CN 106 shown may include a Mobility Management Entity (MME) 162, a Serving Gateway (SGW) 164, and a Packet Data Network (PDN) Gateway (or PGW) 166. While each of the above elements is depicted as part of CN 106, it should be understood that any of these elements may be owned and / or operated by an entity other than a CN operator.

[0050] The MME 162 can connect to each eNodeB 162a, 162b, 162c in RAN104 via the S1 interface and can act as a control node. For example, the MME 162 can be responsible for authenticating users of WTRUs 102a, 102b, 102c, bearer activation / deactivation, selecting a specific serving gateway during the initial attachment of WTRUs 102a, 102b, 102c, etc. The MME 162 can provide control plane functions for handover between RAN104 and other RANs (not shown) employing other radio technologies such as GSM and / or WCDMA.

[0051] The SGW 164 can connect to each eNode B 160a, 160b, and 160c in RAN104 via the S1 interface. The SGW 164 typically routes and forwards user packets to and from WTRUs 102a, 102b, and 102c. The SGW 164 can perform other functions, such as anchoring the user plane during inter-eNode B handover, triggering paging when DL data is available for WTRUs 102a, 102b, and 102c, managing and storing WTRU 102a, 102b, and 102c scenarios, etc.

[0052] The SGW 164 can connect to the PGW 166, which can provide WTRU 102a, 102b, and 102c with access to packet-switched networks (such as the Internet 110) to facilitate communication between WTRU 102a, 102b, and 102c and IP-enabled devices.

[0053] CN 106 can facilitate communication with other networks. For example, CN 106 can provide WTRUs 102a, 102b, and 102c with access to a circuit-switched network (e.g., PSTN 108) to facilitate communication between WTRUs 102a, 102b, and 102c and traditional terrestrial line communication equipment. For example, CN 106 may include, or communicate with, an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server), which acts as an interface between CN 106 and PSTN 108. Furthermore, CN 106 can provide WTRUs 102a, 102b, and 102c with access to other networks 112, which may include other wired and / or wireless networks owned and / or operated by other service providers.

[0054] Despite Figure 1A-1D The WTRU is described as a wireless terminal, but in some representative embodiments, it is envisioned that such a terminal may communicate with a communication network using (e.g., temporarily or permanently) a wired communication interface.

[0055] In a representative embodiment, another network 112 may be a WLAN.

[0056] A WLAN in Infrastructure Basic Services Set (BSS) mode may have an access point (AP) for the BSS and one or more stations (STAs) associated with that AP. The AP may access or interface to a distribution system (DS) or another type of wired / wireless network that carries traffic to and from the BSS. Traffic originating outside the BSS destined for a STA can reach and be delivered to the STA via the AP. Traffic originating from a STA destined for a destination outside the BSS can be sent to the AP for delivery to the appropriate destination. Traffic between STAs within the BSS can be sent via, for example, an AP, where the source STA can send traffic to the AP, and the AP can deliver the traffic to the destination STA. Traffic between STAs within the BSS can be considered and / or referred to as point-to-point traffic. Point-to-point traffic can be sent between the source STA and the destination STA (e.g., directly between the source STA and the destination STA) using a direct link setup (DLS). In some representative embodiments, the DLS may use 802.11e DLS or 802.11z tunneled DLS (TDLS). A WLAN using Standalone BSS (IBSS) mode may not have an access point (AP), and STAs within the IBSS or using the IBSS (e.g., all STAs) can communicate directly with each other. The IBSS communication mode may sometimes be referred to as the "ad-hoc" communication mode in this document.

[0057] When operating in 802.11ac infrastructure mode or a similar mode, the AP can transmit beacons on a fixed channel, such as the primary channel. The primary channel can be of fixed width (e.g., a bandwidth of 20 MHz) or can be dynamically set via signaling. The primary channel can be the operating channel of the BSS and can be used by STAs to establish connections with the AP. In some representative embodiments, Carrier Sense Multiple Access with Collision Avoidance (CSMA / CA) can be implemented, for example in an 802.11 system. With CSMA / CA, all STAs, including the AP (e.g., each STA), can listen on the primary channel. If a particular STA listens / detects and / or determines that the primary channel is busy, that particular STA can back off. A single STA (e.g., only one station) can transmit at any given time within a given BSS.

[0058] High-throughput (HT) STAs can communicate using a 40 MHz wide channel, for example, by combining a primary 20 MHz channel with adjacent or non-adjacent 20 MHz channels to form a 40 MHz wide channel.

[0059] Ultra-high throughput (VHT) STAs can support channels with widths of 20 MHz, 40 MHz, 80 MHz, and / or 160 MHz. 40 MHz and / or 80 MHz channels can be formed by combining consecutive 20 MHz channels. A 160 MHz channel can be formed by combining eight consecutive 20 MHz channels or by combining two non-consecutive 80 MHz channels (this can be referred to as an 80+80 configuration). For the 80+80 configuration, data, after channel coding, can be transmitted via a segmented parser that divides the data into two streams. Each stream can be processed separately using inverse fast Fourier transform (IFFT) and time-domain processing. The streams can be mapped onto the two 80 MHz channels and transmitted by the transmitting STA. At the receiver of the receiving STA, the above operations for the 80+80 configuration can be reversed, and the combined data can be sent to the Media Access Control (MAC).

[0060] 802.11af and 802.11ah support operating modes below 1 GHz. The channel operating bandwidth and carrier in 802.11af and 802.11ah are reduced compared to those used in 802.11n and 802.11ac. 802.11af supports 5 MHz, 10 MHz, and 20 MHz bandwidths in the TV white space (TVWS) spectrum, while 802.11ah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non-TVWS spectrum. According to a representative embodiment, 802.11ah may support Meter Type Control / Machine-Type Communications, such as MTC devices in macro coverage areas. MTC devices may have specific capabilities, such as supporting (e.g., only supporting) specific and / or limited bandwidth functions. MTC devices may include batteries with a battery life exceeding a threshold (e.g., maintaining a very long battery life).

[0061] WLAN systems that can support multiple channels and channel bandwidths (such as 802.11n, 802.11ac, 802.11af, and 802.11ah) include a channel that can be designated as the primary channel. The bandwidth of the primary channel can be equal to the maximum common operating bandwidth supported by all STAs in the BSS. The bandwidth of the primary channel can be set and / or limited by the STA that supports the minimum bandwidth operating mode among all STAs operating in the BSS. Taking 802.11ah as an example, for a STA that supports (e.g., only supports) the 1 MHz mode (e.g., MTC type device), the primary channel bandwidth can be 1 MHz, even if the AP and other STAs in the BSS support 2 MHz, 4 MHz, 8 MHz, 16 MHz, and / or other channel bandwidth operating modes. Carrier Sense and / or Network Assignment Vector (NAV) settings may depend on the status of the primary channel. If the primary channel is busy (e.g., due to a STA (which only supports the 1 MHz operating mode)) transmitting to the AP, the entire available band can be considered busy, even if most of the band remains idle and may be available.

[0062] In the United States, the available frequency band for 802.11ah is 902 MHz to 928 MHz. In South Korea, the available frequency band is 917.5 MHz to 923.5 MHz. In Japan, the available frequency band is 916.5 MHz to 927.5 MHz. The total available bandwidth for 802.11ah is 6 MHz to 26 MHz, depending on the country code.

[0063] Figure 1D This is a system diagram illustrating RAN 113 and CN 115 according to an embodiment. As described above, RAN 113 can communicate with WTRUs 102a, 102b, and 102c via air interface 116 using NR radio technology. RAN 113 can also communicate with CN 115.

[0064] RAN 113 may include gNBs 180a, 180b, and 180c, but it should be understood that RAN 113 may include any number of gNBs while maintaining consistency with the embodiments. gNBs 180a, 180b, and 180c may each include one or more transceivers for communicating with WTRUs 102a, 102b, and 102c via air interface 116. In one embodiment, gNBs 180a, 180b, and 180c may implement MIMO technology. For example, gNBs 180a and 180b may utilize beamforming to transmit signals to and / or receive signals from gNBs 180a, 180b, and 180c. Therefore, for example, gNB 180a may use multiple antennas to transmit radio signals to and / or receive radio signals from WTRU 102a. In embodiments, gNBs 180a, 180b, and 180c can implement carrier aggregation technology. For example, gNB 180a can transmit multiple component carriers to WTRU 102a (not shown). A subset of these component carriers may be located on unlicensed spectrum, while the remaining component carriers may be located on licensed spectrum. In embodiments, gNBs 180a, 180b, and 180c can implement Coordinated Multipoint (CoMP) technology. For example, WTRU 102a can receive coordinated transmissions from gNBs 180a and 180b (and / or gNB 180c).

[0065] WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using transmissions associated with an scalable set of parameters. For example, the OFDM symbol spacing and / or OFDM subcarrier spacing can vary for different transmissions, different cells, and / or different portions of the radio transmission spectrum. WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using subframes or transmission time intervals (TTIs) of various or scalable lengths (e.g., including different numbers of OFDM symbols and / or absolute time lengths of varying durations).

[0066] gNBs 180a, 180b, and 180c can be configured to communicate with WTRUs 102a, 102b, and 102c in standalone and / or non-standalone configurations. In standalone configuration, WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c without also accessing other RANs (e.g., eNode-Bs 160a, 160b, and 160c). In standalone configuration, WTRUs 102a, 102b, and 102c can utilize one or more gNBs 180a, 180b, and 180c as mobility anchors. In standalone configuration, WTRUs 102a, 102b, and 102c can communicate with gNBs 180a, 180b, and 180c using signals in unlicensed frequency bands. In a non-standalone configuration, WTRUs 102a, 102b, and 102c can communicate / connect with gNBs 180a, 180b, and 180c, while also communicating / connecting with another RAN (such as eNode-Bs 160a, 160b, and 160c). For example, WTRUs 102a, 102b, and 102c can implement DC principles to communicate substantially simultaneously with one or more gNBs 180a, 180b, and 180c, as well as one or more eNode-Bs 160a, 160b, and 160c. In a non-standalone configuration, eNode-Bs 160a, 160b, and 160c can act as mobility anchors for WTRUs 102a, 102b, and 102c, while gNBs 180a, 180b, and 180c can provide additional coverage and / or throughput for serving WTRUs 102a, 102b, and 102c.

[0067] Each of gNBs 180a, 180b, and 180c can be associated with a specific cell (not shown in the figure) and can be configured to handle radio resource management decisions, handover decisions, user scheduling in UL and / or DL, network slicing support, dual connectivity, interoperability between NR and E-UTRA, routing user plane data to User Plane Functions (UPF) 184a and 184b, routing control plane information to Access and Mobility Management Functions (AMF) 182a and 182b, etc. Figure 1D As shown, gNB 180a, 180b, and 180c can communicate with each other via the Xn interface.

[0068] Figure 1DThe CN 115 shown may include at least one AMF 182a, 182b, at least one UPF 184a, 184b, at least one Session Management Function (SMF) 183a, 183b, and may include data networks (DN) 185a, 185b. While each of the above elements is depicted as part of the CN 115, it should be understood that any of these elements may be owned and / or operated by an entity other than a CN operator.

[0069] AMF 182a and 182b can connect to one or more of gNBs 180a, 180b, and 180c in RAN 113 via the N2 interface and can act as control nodes. For example, AMF 182a and 182b can be responsible for authenticating users of WTRU 102a, 102b, and 102c, supporting network slicing (e.g., handling different PDU sessions with different requirements), selecting specific SMF 183a and 183b, managing registration areas, terminating NAS signaling, mobility management, etc. AMF 182a and 182b can use network slicing to customize CN support for WTRU 102a, 102b, and 102c based on the service types used by WTRU 102a, 102b, and 102c. For example, different network slices can be established for different use cases, such as services relying on Ultra Reliable Low Latency (URLLC) access, services relying on Enhanced Massive Mobile Broadband (eMBB) access, services for Machine Type Communication (MTC) access, etc. The AMF162 can provide control plane functions for handover between RAN 113 and other RANs (not shown) that employ other radio technologies (such as LTE, LTE-A, LTE-A Pro) and / or non-3GPP access technologies (such as WiFi).

[0070] SMFs 183a and 183b can connect to AMFs 182a and 182b in CN 115 via the N11 interface. SMFs 183a and 183b can also connect to UPFs 184a and 184b in CN 115 via the N4 interface. SMFs 183a and 183b can select and control UPFs 184a and 184b, and configure traffic routing through UPFs 184a and 184b. SMFs 183a and 183b can perform other functions, such as managing and allocating WTRU IP addresses, managing PDU sessions, controlling policy enforcement and QoS, and providing downlink data notifications. PDU session types can be IP-based, non-IP-based, Ethernet-based, etc.

[0071] UPF 184a and 184b can connect to one or more of gNB 180a, 180b, and 180c in RAN 113 via the N3 interface. The N3 interface can provide WTRU 102a, 102b, and 102c with access to packet-switched networks (such as Internet 110) to facilitate communication between WTRU 102a, 102b, and 102c and IP-enabled devices. UPF 184a and 184b can perform other functions such as routing and forwarding packets, enforcing user plane policies, supporting multihomed PDU sessions, handling user plane QoS, buffering downlink packets, providing mobility anchoring, and so on.

[0072] CN 115 can facilitate communication with other networks. For example, CN 115 may include an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server), or may communicate with such an IP gateway, which acts as an interface between CN 115 and PSTN 108. Furthermore, CN 115 can provide WTRUs 102a, 102b, and 102c with access to other networks 112, which may include other wired and / or wireless networks owned and / or operated by other service providers. In one embodiment, WTRUs 102a, 102b, and 102c can be connected to local data networks (DNs) 185a and 185b via UPFs 184a and 184b through their N3 interfaces and the N6 interface between UPFs 184a and 184b and DNs 185a and 185b.

[0073] Given Figure 1A-1D and Figure 1A-1D The functions described herein for one or more of the following can be performed by one or more emulation devices (not shown in the figures): WTRU 102a-d, base station 114a-b, eNode-B 160a-c, MME 162, SGW 164, PGW 166, gNB 180a-c, AMF 182a-ab, UPF184a-b, SMF 183a-b, DN185a-b, and / or any other device(s) described herein. An emulation device can be one or more devices configured to simulate one or more of the functions described herein. For example, an emulation device can be used to test other devices and / or simulate network and / or WTRU functions.

[0074] The simulation device may be designed to perform one or more tests on other devices in a laboratory environment and / or a carrier network environment. For example, the one or more simulation devices may perform one or more or all functions, fully or partially implemented and / or deployed as part of a wired and / or wireless communication network, to test other devices within that communication network. The one or more simulation devices may perform one or more or all functions, temporarily implemented / deployed as part of a wired and / or wireless communication network. The simulation device may be directly coupled to another device for testing, and / or may perform tests using over-the-air wireless communication.

[0075] One or more simulation devices can perform one or more functions without being implemented / deployed as part of a wired and / or wireless communication network. For example, simulation devices can be used in test scenarios within a test laboratory and / or an undeployed (e.g., testing) wired and / or wireless communication network to perform testing on one or more components. One or more simulation devices can be test rigs. Simulation devices can transmit and / or receive data using direct RF coupling and / or wireless communication via RF circuitry (e.g., which may include one or more antennas).

[0076] This paper describes WTRU's approach for implementing online learning for the CSF function in systems using artificial intelligence / machine learning (AI / ML) for channel state feedback (CSF) functionality.

[0077] Wireless devices need to adapt to their local environmental conditions, such as radio propagation environments. For better adaptation, channel state information (CSI) can be known at both the transmitting and receiving nodes. With CSI known at both the transmitting and receiving nodes, the transmission can be appropriately configured and / or received. CSI can be used at the transmitter to select modulation and coding schemes (MCS), layers, precoders, power and / or resource block (RB) allocations, etc., and at the receiver for demodulation (explicitly and / or implicitly). Non-AI / ML algorithms (e.g., channel estimators and / or predictors) can typically estimate CSI. Non-AI / ML algorithms themselves use the output of other non-AI / ML algorithms (e.g., Doppler estimators) to estimate, predict, or provide other channel-related statistics. AI / ML can be used to learn CSI feedback functions (e.g., estimators, predictors, compressors, and / or combinations thereof).

[0078] When learning these CSI feedback functions offline, error-free benchmark real channel data is typically available as a training dataset. Offline training data can be obtained from a limited number of available and / or conventional channel models. These channel models may have limited support for real-world mobility, exclude radio impairments, and / or may not be well-matched to the local environment in which the WTRU resides. To adapt the device to its environment, AI / ML can learn CSF functions optimized for the local environment if online measurements of WTRU- and environment-specific training data are available.

[0079] The input to an AI / ML model can be a pilot (e.g., reference) signal received and / or processed after applying a Fast Fourier Transform (FFT). The model's desired output (also referred to as the "label" and / or "target") may not be directly observable. Instead, a method may be needed to create a high-quality estimate of the channel used as the label. Obtaining high-quality training labels is challenging in operating systems because it may be necessary to minimize the overhead associated with the pilot (e.g., reference) signal. Minimizing the pilot (e.g., reference) signal can optimize throughput and / or other key performance indicators (KPIs). Training an AI / ML-based channel estimator with a conventional estimator and / or conventional pilot signals may not produce a better estimate than a conventional estimator.

[0080] In the example, machine learning (ML) algorithms can be applied to combinations and / or interpolation fusions of the methods described above. For instance, a semi-supervised learning method can use a combination of a small amount of labeled data and a large amount of unlabeled data during training. In this setting, semi-supervised learning falls between unsupervised learning (e.g., utilizing unlabeled training data) and supervised learning (e.g., utilizing only labeled training data).

[0081] Deep learning (DL) refers to a class of machine learning algorithms that utilize artificial neural networks, particularly deep neural networks (DNNs). DNNs are largely inspired by biological systems. A DNN is a special category of ML models inspired by the human brain, where the input can be linearly transformed and passed multiple times through a non-linear activation function. A DNN typically consists of multiple layers. Each layer can consist of a linear transformation and / or a given non-linear activation function. Examples of linear transformations can include: the product of the layer's input and the layer's learnable weights, where learnable biases are added to the resulting product; or a convolution operation between the layer's input and the layer's learnable weights, where learnable biases are added to the resulting convolution.

[0082] A DNN can be trained using training data via a backpropagation algorithm. Backpropagation is the operation of calculating the loss of an AI / ML model (e.g., in the case of supervised learning, the measured difference between the output of the AI / ML model and the target label (e.g., the true benchmark)). Backpropagation may involve calculating the derivative of the calculated loss with respect to the learnable parameters of the AI / ML model (e.g., weights and / or biases, etc.), and then backpropagating these derivatives (e.g., passing the derivatives in reverse order) across the layers of the AI / ML model (e.g., from the output layer (the last layer) to the input layer (the first layer)). The computation may require using the differential chain rule to compute the gradient of each layer with respect to its parameters and / or using the computed gradients to update the parameters of each layer.

[0083] The CSI feedback function can define a series of functions implemented on the WTRU side to enable estimation of Channel State Information (CSI) and / or its transmission to the network. The network can utilize the received CSI feedback to apply link adaptation functions to the WTRU (e.g., appropriate MCS and / or precoding, beam management, power allocation, and / or resource block (RB) allocation, etc.). The CSI feedback function can primarily include CSI estimation. As detailed in 3GPP Release 15, based on the CSI information, the WTRU can generate a CSI report containing measurements detailing the channel state of 5G New Radio (NR), such as Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), Rank Indicator (RI), and / or Layer Indicator (LI), etc.

[0084] Mechanisms and / or frameworks can be enabled for using artificial intelligence / machine learning (AI / ML) based methods at the air interface level, for example, for CSI compression and CSI prediction.

[0085] CSI compression can be defined as follows: the WTRU compresses the CSI estimate within a predefined feedback size (in bits) and / or transmits the compressed CSI estimate to the network. The network can then reconstruct the CSI estimate by decompressing the CSI feedback received from the WTRU. Additionally or alternatively, CSI prediction can define the operation of the WTRU or the network predicting subsequent CSIs based on historical CSI estimates. These two sub-use cases can be implemented within the CSI feedback function and / or after the CSI estimation operation.

[0086] AI / ML operations for CSI feedback functionality can be based on models (e.g., neural networks) trained on large amounts of data under different scenarios and / or conditions. For a given function, several models may exist (e.g., each trained and / or adapted to different networks and / or WTRU conditions and / or characteristics). Once the AI / ML model is trained, it can be deployed for online inference. After deployment, AI / ML model monitoring can be performed on a real network, as the current network and / or WTRU conditions may change from the scenarios and / or conditions in which the model was trained and / or tested. AI / ML model training and / or monitoring of the CSI feedback functionality can be performed collaboratively between WTRU, the network, and / or both.

[0087] After deploying an AI / ML model for online inference, its performance may degrade over time due to a concept known as "model drift." The deployed AI / ML model continuously receives new data to make predictions based on it. However, this data may have a different probability distribution than the data used for training. Therefore, using the original AI / ML model with the new data distribution can lead to a decline in model performance. AI / ML model drift can be associated with a decrease in model performance over time. This performance degradation can cause the model to give poorer predictions. AI / ML model drift can be categorized as concept drift or data drift.

[0088] Conceptual drift can occur when the functional relationship between the input and output of an AI / ML model changes. The functional relationship may have changed, but the model may not be aware of this change. Therefore, the model may no longer retain the learned parameters. Data drift can occur when the distribution of the input to an AI / ML model changes. The performance of the AI / ML model may degrade because the model has not received enough data and / or has been trained sufficiently. Data drift can also occur if the data received by the AI / ML model during online inference contains features that were not present during offline training.

[0089] During online inference of an AI / ML model, the WTRU and / or network AI / ML endpoints (e.g., gNB, cloud servers, and / or edge computing servers, etc.) can detect that the AI / ML model's performance has degraded. When the WTRU first enters a new geographic area, the AI / ML model may not be well trained for that new local environment. In this case, the WTRU can request support to update the AI / ML model. One solution is that the WTRU can retrain the AI / ML model online with the assistance of another node in the network (e.g., gNB, another WTRU using a side link (SL) and / or WTRU-to-WTRU direct communication, etc.). Although the proposed solution can be applied to any CSI feedback function (e.g., CSI estimation, CSI prediction, CSI compression, and / or combinations thereof, etc.), by way of example, the remainder of this disclosure focuses here on CSI estimation. In the context of channel estimation, the solution focuses on providing steps that allow the WTRU to perform online training of the AI / ML model.

[0090] The WTRU and / or network can detect the need to initiate online training of the channel estimation algorithm. In the example, the AI / ML model drift detection mechanism in the WTRU can indicate that the AI / ML model has drifted and / or is drifting. The WTRU can enter geographic areas and / or cells where the channel estimation algorithm has not been previously trained (e.g., the WTRU compares Global Positioning System (GPS) coordinates to measure the distance to the previously trained area). If the distance to the previously trained area exceeds a threshold provided by the network, the WTRU can signal to the network that the threshold has been exceeded. The WTRU may also be able to signal its distance and / or current location.

[0091] If the WTRU supports an AI / ML-based CSI estimator using data symbols (in addition to Channel State Information Reference Signals (CSI-RS)), the WTRU capability message can indicate this support. If the WTRU supports an AI / ML-based CSI estimator using data symbols (in addition to CSI-RS), the WTRU can request a radio resource allocation for training, which uses pseudo-random data (PRD) training allocation instead of data resource elements (REs).

[0092] In addition to PRD training allocation, the WTRU can also request and / or use Learning Reference Signals (L-RS) training allocation. The WTRU can generate a CSI report based on the CSI-RS. The network can use previous CSI report information to select a precoder for the PRD training allocation. The network can analyze the CSI report. The network can determine whether to perform a PRD training allocation for the WTRU. The network can confirm the request for a PRD training allocation, indicating rejection or acceptance. If accepted, the network can send the key, seed, and / or other information required for locally generating pseudo-random bits and / or symbols to the WTRU. After receiving this information, the WTRU can now be ready to receive the PRD training allocation. The network can instruct the WTRU that the WTRU can precode the PRD according to the CSI report from the WTRU's last report.

[0093] The WTRU can assume that the demodulation reference signal (DMRS) used for PRD training allocation uses a precoder consistent with its most recent CSI report. If the WTRU supports blind detection of PRD training allocation, the WTRU capability message can indicate this support. The network can make PRD training allocations for the WTRU without indicating that the allocation is a PRD training allocation. Aperiodic PRD training allocations can include PRD training indications in downlink control information (DCI) and / or other downlink (DL) control channels. Semi-persistent PRD training allocations can be configured without WTRU-specific (and / or WTRU group-specific) signaling (e.g., Radio Resource Control (RRC) and / or MAC-CE). Periodic and / or semi-persistent PRD allocations can be indicated as opportunistic. If a PRD allocation is opportunistic, the WTRU can determine whether the allocation contains a PRD before using the allocation for training.

[0094] A PRD (Programming Request for Responsibility) can indicate the amount of training bandwidth (BW) and / or baseband (RB). The network can determine the size of the training allocation (e.g., how large). The request can indicate channel statistical measurements, such as delay spread and / or Doppler shift. The request can indicate channel diversification (e.g., increased delay spread or Doppler shift). The network can then determine the magnitude (e.g., how much) of any changes to be added to the channel (e.g., a time-varying finite impulse response (FIR) filter can enhance the channel).

[0095] This request can indicate the priority of the request. Here, if the channel estimator's performance is below a certain threshold, the WTRU can indicate a high priority. The WTRU can receive this performance threshold from the network (e.g., gNB). The network can indicate in the System Information Block (SIB) and / or RRC that the WTRU may require estimation model drift and / or Normalized Mean Square Error (NMSE) above the threshold for the channel estimation to make a priority request.

[0096] The network can decide to cover the PRD training assignment with data RE. If the priority is low, the network can also decide to reduce or not increase the training assignment. If the request is for predictive maintenance (e.g., to compare the quality of an AI / ML-based CSI estimator with the CSI estimated by a reference CSI estimator using L-RS), the WTRU can indicate a low priority. The WTRU supporting PRD training can also request and / or use an existing L-RS training assignment to measure the performance of the AI / ML estimator. If the comparison indicates that the AI / ML performance has degraded compared to a threshold, the WTRU can signal to the network to increase the priority of the training assignment.

[0097] WTRUs can use non-specific training assignments. These assignments can be immediate (notified by signaling in the DCI and / or similar control channels) and / or semi-static, where training assignment information is indicated in the SIB and / or other broadcast messages. Non-specific training assignments can be non-specific to any WTRU. Therefore, multiple WTRUs can use the same training assignment simultaneously. Non-specific training assignments can include PRD training assignments and / or L-RS training assignments. In PRD training assignments, data REs can be populated with seemingly random data that can be used as input to an AI / ML-based CSI estimator (e.g., a trainable channel estimator). The same seemingly random data can be a known sequence. The same seemingly random data can be used for high-quality channel estimation as a label for a trainable (e.g., AI / ML) channel estimator using regular CSI-RS as input.

[0098] Non-specific training assignments can be configured as semi-persistent, periodic, semi-periodic, and / or periodic. Figure 2 This is an example of a Type 1 training assignment 200 with a PRD and CSI-RS. A Type 1 PRD training assignment can include a PRD 204 (e.g., symbols) in a RE that typically carries data. Furthermore, the RE used for DMRS can be used for PRD DRMS ​​208. The RE used for DMRS can be pre-encoded in the same manner as PRD 204. In this method, as... Figure 2As depicted, the allocated RBs (e.g., resource block groups (RBGs)) may not carry any user data. Data REs can alternatively be filled with PRD RE 204. CSI-RS 212, control channels such as PDCCH 216, and / or other physical (PHY) channels may be present. PRD RE 204 can be derived from a key- and / or seed-based random number source. Therefore, PRD RE 204 can be preprocessed and treated as a pilot sequence for high-quality channel estimation, serving as a label for a trainable channel estimator using conventional CSI-RS 212 as input.

[0099] The same PRD and / or, for example, PRD DMRS RE (but without preprocessing based on known PRD sequences) can then be used as input to the AI / ML model during training. Any current CSI-RS can also be used as input to the AI / ML model during training. In this way, the model can learn to use data RE as input in addition to CSI-RS. The model can learn to generate high-quality labels simultaneously using the same PRD data and / or signals. Since the network does not need to use the precoder indicated by the WTRU CSI report, WTRU can use only the CSI-RS as input to a second trainable channel estimator. This second trainable channel estimator may not use data RE as input during interface and / or may not use PRD as input during training.

[0100] Figure 3 This is an example of a Type 2 training allocation 300 with PRD, CSI-RS, and / or synchronous data scheduling. The Type 2 PRD training allocation includes PRD 304 (e.g., symbols) and / or data REs. A Type 2 PRD training allocation can make a portion of the RE available for user data 308. A common DRMS ​​312 is used for both data 308 and / or PRD 304. The same antenna port and / or the same precoder are used for data 308, PRD 304, and / or DMRS 312. In this method, as... Figure 3 As depicted, some REs in the training allocation can be populated with PRD RE 304, enabling the creation of high-quality channel estimates as labels for a trainable channel estimator using regular CSI-RS 316 and / or data RE 308 as input. CSI-RS 316, control channels such as PDCCH 320, and / or other physical (PHY) channels may be present. Note that if PRD 304 uses the same modulation as the data, the input to the AI / ML model can include data 308 and / or PRD 304. It should also be noted that high-quality CSI estimates used for CSI labeling can also be used as channel estimates for CSI reporting and / or demodulation.

[0101] Type 2a PRD training allocation can refer to a training allocation that occupies the same (e.g., identical) time slot RB and a WTRU data allocation. WTRUs capable of using Type 2a PRD training allocations for data can be scheduled for simultaneous training and / or data acquisition. These WTRUs can learn patterns of PRDs and / or WTRU-specific REs from SIB, RRC, and / or other broadcast information. WTRUs capable of training using Type 2a PRD training allocations (such WTRUs must be part of a WTRU group for which another WTRU is acquiring data) can learn PRD patterns from SIB, RRC, and / or other broadcast information. Type 2a allocation information in the SIB can indicate that a WTRU-specific RE can be populated with a PRD when no data is available in a Type 2a allocation. If indicated in the SIB, a WTRU can detect the presence of a PRD in a WTRU-specific RE. WTRUs can use the PRD as additional input for improved CSI labels and / or AI / ML estimators.

[0102] Type 2b non-specific training assignments (training assignments and / or WTRU data assignments can use different but overlapping time slots RB) can utilize the same process as Type 2a. However, from a training perspective, the WTRU should know which RBs in the training assignment contain WTRU-specific REs. A WTRU capable of training with a Type 2b training assignment can indicate this capability to the network. Therefore, the WTRU indicates to the network that placing it in a group using a Type 2b training assignment may not incorrectly incorporate user data REs into the AI / ML training data.

[0103] WTRU can also be configured for blind detection. WTRUs assigned using type 2 training can use REs in each RB and / or RBG to determine whether the RB and / or RBG might carry user data and / or exclude REs carrying user data from AI / ML training.

[0104] WTRUs can also be configured for control channel indication (e.g., PDCCH). Allocations carrying user data can also carry information detectable (at least) by a WTRU with type 2b capability, indicating the RBs used in the data allocation. WTRUs using trained allocations can exclude REs carrying user data from AI / ML training.

[0105] From a data reception perspective, the WTRU should know which Responsible Blocks (RBs) in the data allocation contain the PRD. A WTRU capable of receiving data in a Type 2b PRD training allocation can indicate to the network that it has identified a semi-persistent training allocation. In other words, the WTRU can indicate that it knows the allocation information for the indicated semi-persistent training allocation. The network can then allow simultaneous data and / or training allocations to occur for the WTRU that has made such an indication. When the WTRU receives a data allocation, it can identify the RBs that intersect with the training allocation. Based on the training allocation information, the WTRU can determine which Receiver Entities (REs) from the intersecting RBs can actually be carrying user data. The WTRU can then select those REs accordingly for data reception. If the training allocation is opportunistic (e.g., also determined from the training allocation information), the WTRU can also check REs that similarly carry the PRD. Opportunistic training allocations can enable the WTRU to detect PRD training allocations. The WTRU can perform this check to determine whether the RE carries user data and / or the PRD and / or select those REs accordingly for data reception.

[0106] WTRUs can receive broadcast information to learn the current PRD training allocation schedule for WTRUs and / or WTRU groups. WTRUs can read broadcast information (e.g., SIBs and / or RRCs) to learn semi-static scheduling of PRD training allocations for Type 1, Type 2a, and / or Type 2b. Allocation information can be similar to other semi-static allocations (e.g., start SFN and slots, RBs and / or RBGs, periods and / or frequency hopping, etc.) and / or PRD-specific information. PRD training allocations may include resource allocation information and / or PRD power offset information (e.g., power offset relative to CSI-RS).

[0107] PRD training assignments may include PRD number generator status and / or seed and / or corresponding system frame number (SFN). These can be included in the PRD training assignment to allow the WTRU to learn the random symbol sequence used in the PRD RE. The PRD training assignment may include a modulation type indicator. This indicator may indicate a sequence of different modulation orders. For example, if the WTRU reads the modulation indicator {0,1,0,2,3,0,0}, then the first, third, sixth, and seventh RBs are Quadrature Phase Shift Keying (QPSK), the second RB is 16 Quadrature Amplitude Modulation (QAM), the fourth RB is 64QAM, and the fifth RB is 256QAM. Semi-periodic PRD training assignments may include a 'chance' indicator. The chance indicator can enable the WTRU to detect PRD training assignments. This chance indicator may exist as an information bit in the SIB. If indicated, the WTRU may not consider the assignment a PRD training assignment. Instead, the WTRU may detect whether the assignment is a PRD training assignment and / or use the assignment for training. Detection may include correlation with known PRD symbol sequences. Opportunity indicators can be applied to Type 1 and / or Type 2 semi-static training assignments.

[0108] If multiple PRDs are used for training assignments, WTRUs can be assigned to WTRU PRD groups. Directional systems may require this grouping. Different ports can be used for different directions (e.g., WTRU groups). Furthermore, non-directional systems may also require this grouping. For example, WTRUs can be grouped based on the possible precoders to be used. The network can learn the most likely precoders to use from the WTRU CSI report. The correlation between the CSI report and / or precoder selection helps the trainable channel estimator utilize DMRS and / or data RE. WTRUs can read the schedule of the WTRU PRD group to which they belong.

[0109] Figure 4This is an example diagram of online data labeling and / or training for a CSI estimator based on artificial intelligence / machine learning (AI / ML). WTRU 404 can receive PRD data 408 and / or apply preprocessing 412. In this case, preprocessing 412 can include, but is not limited to, extracting the PRD RE from the received PRD signal using PRD resource allocation. Preprocessing 412 can include dividing by the PRD symbol of a known sequence or multiplying by the conjugate of the PRD symbol of a known sequence. Preprocessing 412 can include 2D finite impulse response (FIR) filtering options for the RE carrying the PRD. WTRU 404 can use the preprocessed data 412 as input to a channel estimator 416 (e.g., a high-quality channel estimator). WTRU 404 can use the preprocessed PRD 412 to generate a good quality label 420 (e.g., a CSI label) for the CSI via the channel estimator 416 (e.g., a high-quality channel estimator), such as... Figure 4 As shown in the image.

[0110] WTRU 404 can use PRD 408 as input to a trainable channel estimator 424. In this case, preprocessing 428 may include, but is not limited to, extracting PRD REs from the received PRD signal 408 using PRD resource allocation. WTRU 404 can use the preprocessed PRD 428 as input to a trainable channel estimator 424 (e.g., an AI / ML model, an AI / ML channel estimator, etc.) and / or generate output CSI 432, such as... Figure 4 As shown in the image.

[0111] WTRU 404 can train a trainable channel estimator (e.g., AI / ML model, AI / ML channel estimator, etc.) based on the generated CSI label 420 and / or the output CSI 432. Figure 4As shown in the diagram. WTRU 404 can monitor training progress by, for example, measuring the rate of change of the loss function 436 and / or comparing it with a threshold. For example, the training of a trainable channel estimator 424 (e.g., an AI / ML model, an AI / ML channel estimator, etc.) can be based on CSI labels 420 and the output CSI 432 by comparing them with a loss function (e.g., the measured difference between CSI labels 420 and the output CSI 432) (e.g., inputting them into the loss function) and calculating the loss 436, and using backpropagation to update the model's parameters (e.g., layer weights and / or biases). WTRU 404 can then report to network 440 (e.g., gNB) that training is sufficient. WTRU 404 can perform online inference using the trained channel estimator 424 (e.g., an AI / ML model, an AI / ML channel estimator, etc.). The input to the trained channel estimator 412 can be regular CSI-RS 420 received from network 412 (e.g., gNB). The output can be a CSI432 output associated with a received conventional CSI-RS 444.

[0112] During online inference of AI / ML models, WTRUs and / or network AI / ML endpoints (e.g., gNBs, cloud servers, and / or edge computing servers) may need to monitor the performance of the AI / ML models. In the context of channel estimation used for probing, the following process focuses on providing steps that allow WTRUs to perform performance monitoring of AI / ML models using PRDs.

[0113] Figure 5 This is an example graph illustrating performance monitoring of an AI / ML-based CSI estimator using PRD. The WTRU 504 can request and / or monitor the performance of channel estimation algorithms, such as... Figure 5 As shown in the diagram. WTRU 504 can perform performance monitoring using PRD 508. WTRU 504 can perform performance monitoring using Type 1, Type 2a, and / or Type 2b PRD scheduling. The monitoring request can indicate the amount of test bandwidth (BW), bandwidth portion (BWP), and / or RB. The network can use this request to determine how large the test allocation should be. WTRU 504 can receive performance thresholds from network 540 (e.g., gNB). The network (e.g., gNB) can indicate the thresholds in SIB and / or RRC. To continue using the same AI / ML model, the estimated NMSE of the channel estimation may not exceed this threshold. WTRU 504 can use non-specific test allocations. These allocations can be immediate (e.g., notified by signaling in DCI and / or similar control channels) and / or semi-static, where test allocation information is indicated in SIB and / or other broadcast messages.

[0114] WTRU 504 can receive PRD 508 and / or apply preprocessing 512. WTRU 504 can use the preprocessed data 512 as input to channel estimator 516 (e.g., a high-quality channel estimator). When WTRU 504 uses PRD 508 to generate CSI tag 520, preprocessing 512 can include extracting a regular PRD RE from the received PRD signal using PRD resource allocation. Preprocessing 512 can include dividing by a known sequence of PRD symbols or multiplying by the conjugate of a known sequence of PRD symbols. Preprocessing 512 can include a 2D FIR filtering option for the RE carrying the PRD. WTRU 504 can use the preprocessed PRD 512 to generate a good-quality tag 520 (e.g., a CSI tag) for the CSI via channel estimator 516 (e.g., a high-quality channel estimator), such as... Figure 5 As shown.

[0115] WTRU 504 can use PRD 508 as input to a trainable channel estimator 524 (e.g., an AI / ML model, AI / ML channel estimator, etc.). In this case, preprocessing 528 may include, but is not limited to, extracting PRD REs from the received PRD signal 508 using PRD resource allocation. WTRU 504 can use the preprocessed PRD 528 as input to a trainable channel estimator 524 (e.g., an AI / ML model, AI / ML channel estimator, etc.) and / or generate the CSI 532 of the output, such as... Figure 5 As shown in the image.

[0116] WTRU 504 can calculate the error 536 between the CSI label 520 and the output CSI 532. If the error 536 is below a threshold, WTRU 504 can report the result to the network 540 (e.g., gNB) and / or can continue using the same trainable channel estimator 524 (e.g., AI / ML model, AI / ML channel estimator, etc.). If the error 536 exceeds the threshold, WTRU 504 can report the result to the network 540 (e.g., gNB), can fall back to a traditional non-AI / ML channel estimator, and / or can request online retraining of the AI / ML model by requesting a training assignment.

[0117] During online inference of AI / ML models, the WTRU or network AI / ML endpoints (e.g., gNB, cloud servers, and / or edge computing servers) can monitor the performance of the AI / ML model. In the context of channel estimation for probing, the following process focuses on providing steps that allow the WTRU to perform performance monitoring of the AI / ML model using L-RS.

[0118] Figure 6This is an example graph illustrating performance monitoring of an AI / ML-based CSI estimator using a learned reference signal (L-RS). The WTRU 604 can request and / or monitor the performance of the channel estimation algorithm, such as... Figure 6 As shown in the diagram. WTRU 604 can use L-RS 608 for performance monitoring. L-RS 608 is an online learning RS that supports (e.g., in the case of conventional CSI-RS 610) semi-persistent, periodic, and / or aperiodic scheduling. WTRU 604 can use Type 1, Type 2a, and / or Type 2b L-RS scheduling for performance monitoring. Monitoring requests can indicate the number of test BWs, BWPs, and / or RBs. The network can use this request to determine how large and / or how frequently the test assignments should be. WTRU 604 can receive performance thresholds from network 640 (e.g., gNB). Network 640 (e.g., gNB) can indicate in the SIB and / or RRC that the estimated NMSE of the channel estimation is below the threshold to continue using the same AI / ML model. WTRU 604 can use non-specific test assignments. These assignments can be immediate (e.g., notified by signaling in the DCI and / or similar control channels) and / or semi-static, where test assignment information is indicated in the SIB and / or other broadcast messages.

[0119] WTRU 604 can receive conventional CSI-RS 610 and / or apply preprocessing 612. WTRU 604 can use the preprocessed CSI-RS 612 as input to channel estimator 616 (e.g., a high-quality channel estimator). When WTRU 604 uses CSI-RS 610 to generate CSI tags 620, preprocessing 612 can include extracting the conventional CSI-RS 610 received on the allocated resources. Preprocessing 612 can include dividing by a known CSI-RS 610 symbol or multiplying by the conjugate of a known CSI-RS 610 symbol. Preprocessing 612 can include a 2DFIR filtering option for the RE carrying CSI-RS 610. WTRU 604 can apply out-of-distribution (OOD) detection tests to the extracted CSI-RS 610.

[0120] WTRU 604 can receive L-RS 608 and / or apply preprocessing 612. WTRU 604 can use the preprocessed data 612 as input to channel estimator 616 (e.g., a high-quality channel estimator). When WTRU 604 generates CSI tag 620 using L-RS 608, preprocessing 612 can include extracting the L-RS 608 received on the allocated resources. Preprocessing 612 can include dividing by a known L-RS 608 symbol or multiplying by the conjugate of a known L-RS 608 symbol. Preprocessing 612 and / or 2D FIR filtering options are available for REs carrying L-RS 608. WTRU 604 can apply an OOD detection test to the extracted L-RS 608.

[0121] WTRU 604 can apply preprocessing 628 to conventional CSI-RS 610. WTRU 604 can use the preprocessed CSI-RS 628 as input to a trainable channel estimator 624 (e.g., an AI / ML model, AI / ML channel estimator, etc.) to generate an output CSI 632. When WTRU 604 uses CSI-RS 608 to generate the output CSI 632, preprocessing 628 can include extracting conventional CSI-RS 610 received on allocated resources. Preprocessing 628 can include dividing by a known CSI-RS 610 symbol or multiplying by the conjugate of a known CSI-RS 610 symbol. Preprocessing 628 can include a 2DFIR filtering option for the RE carrying CSI-RS 610. WTRU 604 can apply OOD detection tests to the extracted CSI-RS 610. The WTRU 604 can use pre-processed standard CSI-RS 628 as input to a trainable channel estimator 624 (e.g., an AI / ML model, AI / ML channel estimator, etc.). The WTRU 604 can use the trainable channel estimator 624 to generate the output CSI 632, such as... Figure 6 As shown in the image.

[0122] WTRU 604 can calculate the error 636 between CSI label 620 and / or the output CSI 632. If the error 636 is below a threshold, WTRU 604 can report the result to network 640 (e.g., gNB) and / or continue using the same AI / ML model (e.g., AI / ML channel estimator, etc.). If the error exceeds the threshold, WTRU can report the result to network 640 (e.g., gNB), can fall back to a traditional non-AI / ML channel estimator, and / or can request online retraining of the AI / ML model by requesting a training assignment.

[0123] The following describes a method for online learning of the CSI estimator. The WTRU and / or the network can detect that the performance of the AI / ML model has degraded. When a WTRU first enters a new area, the AI / ML model may not be well trained for that new local environment. In this case, the WTRU can request support to update the AI / ML model. In the example, the WTRU can retrain the AI / ML model online with the assistance of another node in the network (e.g., a gNB, another WTRU using a side-link (SL) and / or WTRU-to-WTRU direct communication, etc.). Another example provides the WTRU retraining the AI / ML model online with the assistance of another node in the network (e.g., a gNB, another WTRU using an SL and / or WTRU-to-WTRU direct communication, etc.). This context focuses on providing the steps and / or details that allow the WTRU to perform online training of the AI / ML model.

[0124] When the WTRU is configured to perform online learning of an AI / ML model for CSI estimation of channel sounding, the WTRU can transmit key parameters related to the AI / ML capabilities to the network. These parameters may include the WTRU's use of PRD and / or other training signals depending on its capabilities. Capability messages may include the use of Type 1 training assignments in non-opportunistic mode and / or Type 1 training assignments in opportunistic mode.

[0125] Opportunistic mode may occur when WTRU detects an assignment but does not recognize it as a PRD training assignment. Supporting opportunistic mode means that WTRU can detect when a PRD training assignment is (or not) replaced by data. Furthermore, supporting opportunistic mode means that if WTRU determines that the PRD has been replaced by data, WTRU can choose not to use the PRD assignment for training.

[0126] Opportunistic pattern capabilities may also include training using a type 2a PRD training allocation without an extended PRD. An extended PRD can refer to the possibility that an RE reserved for data during type 2a PRD training allocation can be alternatively replaced by an extension of the PRD (e.g., if no data is available for transfer in those REs, they may contain a PRD). If the WTRU only supports type 2a without extended L-RS, the WTRU can simply ignore the extended PRD (and no testing against the extended PRD is required).

[0127] Opportunistic pattern capabilities may also include the use of type 2a training assignments for training with extended PRDs. Supporting type 2a PRD training assignments with extended PRDs could mean that WTRU detects when PRD expansion occurs and / or includes the extended PRD in high-quality label generation.

[0128] Opportunistic pattern capabilities may also include the use of type 2a training assignments for data reception. Supporting type 2a PRD training assignments for data reception instructs the network that it can use type 2a PRD training assignments to receive data.

[0129] Opportunistic mode capabilities may also include using type 2b training assignments for training without an extended PRD; using type 2b training assignments for training with an extended PRD; using type 2b training assignments for data reception; using type 2a training assignments in opportunistic modes; and / or using type 2b training assignments in opportunistic modes.

[0130] Additionally or alternatively, the network may be configured with other key parameters related to AI / ML functionality via higher-level signaling (e.g., RRC signaling and / or SIB, etc.). This configuration may include informational elements and / or parameters for PRD training assignment scheduling. For example, these elements and parameters may include that the PRD is an online learning training signal that supports (as in the case of conventional CSI-RS) semi-persistent, periodic, and / or aperiodic scheduling. PRD training assignments may be non-specific to any WTRU (or non-specific to any WTRU within one or more WTRU-L-RS groups indicating training assignments). PRDs may be configured with higher power and / or higher time-frequency density than conventional CSI-RS, and / or may be indicated in training assignments as an offset from CSI-RS power.

[0131] Additional configuration for the WTRU used to perform online learning may include multiplexing the WTRU-specific RE with the PRD in a Type 2 training allocation. In the example, the PRD may be multiplexed simultaneously with data transmitted in the New Radio (NR) Physical Downlink Shared Channel (PDSCH) and the DRMS ​​RE, as well as other RSs (e.g., CSI-RS, etc.). The parameter R included in the training allocation information... PRD The ratio of REs used for PRDs in the training allocation can be determined. PRD It can be selected from a finite set of numbers (e.g., a list) and / or can be referenced as an index of that list. (One or more) Additional parameters O PRD and / or N PRD The starting RE offset and / or the number of PRDs to be inserted in the type 2 allocation can be determined. For example, if the REs are listed (e.g., assigned a sequence number), then O PRD It can indicate the first RE carrying the PRD; R PRD It can be determined how many REs to skip until the next PRD is inserted; and / or N. PRDIt can be determined how many PRDs were inserted (and / or the RE numbers for which PRDs will no longer be inserted). The sequence number list may omit a pre-defined set of REs, such as DMRS. The pre-defined list can be part of the training assignment information.

[0132] WTRUs can be configured with additional elements and / or parameters for online learning, including but not limited to: PRD resource mapping (e.g., number of antenna ports, PRD mode, etc.); high-quality channel estimator: L-RS modulation mode; grouping of WTRUs for different PRD opportunities; size of training allocation; channel resource configuration; set of available RBs; thresholds for ML performance monitoring and associated priority levels; and / or AI / ML model training parameters (e.g., loss function, training performance threshold, etc.).

[0133] For online training initiated by the WTRU, the WTRU can transmit an online training request message to the network AI / ML endpoint nodes. This training request can indicate the amount of training bounding winds (BWs) and / or repetitive loads (RBs). The network can use this request to determine how large the training allocation should be. The training request can also indicate channel statistical measurements, such as delay spread and / or Doppler shift; or channel diversification requests (e.g., increased delay spread and / or Doppler shift). The network can use this request to determine how much (if any) variation to add to the channel (e.g., using a time-varying FIR filter to enhance the channel).

[0134] Training requests can include a priority. Specifically, if the channel estimator's performance is below a certain threshold, the WTRU can indicate a high priority. The WTRU can receive the performance threshold from the network (e.g., the network can indicate in the SIB and / or RRC that the estimated NMSE of the channel estimator exceeds the threshold required for the WTRU to make a priority request). If the priority is low, the network can decide not to cover the data RE with training allocations or to increase the training allocation. If the request is for predictive maintenance (e.g., for comparing the quality of the AIML CSI estimator with the estimated CSI of a reference CSI estimator using a PRD), the WTRU can indicate a low priority. If the comparison indicates that the AIML performance has degraded compared to the threshold, the WTRU can signal to the network to notify a new message to increase the priority of the training allocation.

[0135] Training requests may include online training response messages transmitted by network AI / ML endpoint nodes as a response to online training requests transmitted by WTRUs. These messages may include, but are not limited to: the location of the RB and / or PRB transmitting the PRD; the allocated online training duration and start time; and / or the allocated online training L-RS mode and / or density, etc. Online training request messages from WTRUs may be transmitted on the Physical Uplink Control Channel (PUCCH), Contention-Free Physical Random Access Channel (PRACH), PUSCH, Media Access Control Element (MAC-CE) RRC messages, and / or Non-Access Stratum (NAS) messages.

[0136] This paper discloses a process for online training in which WTRU can train an AI / ML model for CSI estimation of probes. The stages and / or steps of the AI / ML process include applying the AI / ML model (e.g., CSI estimation) and / or input data (e.g., CSI-RS with preprocessing and / or PRD with preprocessing, etc.).

[0137] When the WTRU receives a PRD, the WTRU can apply preprocessing. Preprocessing may include extracting the PRD RE from the received PRD signal using PRD resource allocation.

[0138] Regarding AI / ML models, various AI / ML models have been proposed in the literature for CSI estimation, among which the Residual Channel Estimation Network (ReEsNet) may be a good model in terms of performance-complexity tradeoffs. ReEsNet can be based on convolutional layers. The next step may involve output data, where the full channel estimate (e.g., channel impulse response) is correlated with the input (e.g., the extracted conventional CSI-RS).

[0139] Training can be performed online in a supervised manner. The features trained can be the extracted PRD REs. The labels can be the target (e.g., true) full-channel estimates obtained using the same PRD as input data (but after applying additional preprocessing). Preprocessing in this case can include extracting the PRD REs from the received PRD signal using PRD resource allocation; dividing by the PRD symbol of a known sequence (or multiplying by its conjugate); and / or 2D FIR filtering options on the REs carrying the PRD.

[0140] The WTRU can use preprocessed PRDs to power a channel estimator that uses the high-quality and / or quantity of PRDs to generate high-quality labels (e.g., CSI labels) for an AI / ML CSI estimator. The WTRU can then train a trainable (e.g., AI / ML) channel estimator based on the generated CSI labels and / or the output CSIs. The WTRU can then monitor training progress (e.g., by measuring the rate of change of the loss function, comparing it to a threshold, and / or reporting to the network (e.g., gNB) that training is sufficient).

[0141] Next, the WTRU can use a trained channel estimator for online inference. The input to the trained channel estimator can include regular CSI-RS received from the network (e.g., gNB). The input to the trained channel estimator can also include user data and DMRS received from the network (e.g., gNB). The output can be an output CSI estimate.

[0142] The following describes a method for monitoring AI / ML model performance using a learned reference signal. When the WTRU performs drift detection on the AI / ML model estimated by CSI for channel sounding, the network can configure key parameters via higher-level signaling (e.g., RRC signaling and / or SIB, etc.). This configuration may include error thresholds and / or PRD test allocation scheduling.

[0143] This document describes the steps and / or procedures for WTRU to perform AI / ML model performance monitoring for CSI estimation using PRD. WTRU can request and / or monitor the performance of channel estimation algorithms, such as... Figure 5 As shown in the diagram, WTRU can be configured to use PRDs for performance monitoring. WTRUs can use type 1, type 2a, and / or type 2b PRD scheduling for performance monitoring.

[0144] Monitoring requests can indicate the number of test BWs, BWPs, and / or RBs. The network can use this request to determine how large the test allocation should be. WTRUs can receive performance thresholds from the network (e.g., gNBs). WTRUs can indicate in SIBs and / or RRCs that the estimated NMSE of the channel estimation is below a threshold to continue using the same AI / ML model. WTRUs can be configured to use non-specific test allocations. These allocations can be immediate (e.g., signaled in DCIs and / or similar control channels) and / or semi-static, where test allocation information is indicated in SIBs and / or other broadcast messages.

[0145] The WTRU can receive standard CSI-RS and / or apply preprocessing. The WTRU can use the PRD to generate CSI tags. Preprocessing may include extracting the PRD RE received on the allocated resources. Preprocessing may include dividing by the PRD symbol of a known sequence or multiplying by the conjugate of the PRD symbol of a known sequence. Preprocessing may include a 2D FIR filtering option for the RE carrying the PRD.

[0146] like Figure 4 and Figure 5 As shown, WTRU can use a pre-processed PRD to generate a good quality label for CSI (e.g., a CSI label).

[0147] WTRU can use PRD as input to a channel estimator. Preprocessing in this case involves extracting the PRD RE from the received PRD signal using PRD resource allocation. WTRU can use the preprocessed PRD as input to a trainable channel estimator and / or generate the CSI of the output, such as... Figure 4 and Figure 5 As shown in the image.

[0148] The WTRU can calculate the error between the CSI label and / or the output CSI. If the error is below a threshold, the WTRU can report the result to the network (e.g., gNB) and / or continue using the same AI / ML model (e.g., AI / ML channel estimator, etc.). If the error is above the threshold, the WTRU can report the result to the network (e.g., gNB) and / or can request online retraining of the AI / ML model (e.g., AI / ML channel estimator, etc.).

[0149] When the WTRU can perform drift detection using an AI / ML model for CSI estimation for channel sounding with L-RS, the network can configure key parameters via higher-level signaling (e.g., RRC signaling and / or SIB, etc.). This configuration may include error thresholds and / or L-RS test allocation scheduling.

[0150] The steps and / or procedures for the WTRU to perform AI / ML model performance monitoring for CSI estimation can begin when the WTRU requests and / or monitors the performance of the channel estimation algorithm. The WTRU can use L-RS for performance monitoring, such as... Figure 6 As shown in the diagram, the L-RS is an online learning RS (e.g., a regular CSI-RS) that supports semi-persistent, periodic, and / or aperiodic scheduling. WTRU can then use type 1, type 2a, and / or type 2b L-RS scheduling for performance monitoring.

[0151] Monitoring requests can indicate the number of test BWs, BWPs, and / or RBs. The network can use this request to determine how large the test allocation should be. The WTRU can receive performance thresholds from the network (e.g., gNB). The WTRU can indicate in the SIB and / or RRC that the estimated NMSE of the channel estimation is below a threshold to continue using the same AI / ML model. The WTRU can use non-specific test allocations. These allocations can be immediate (e.g., signaled in the DCI and / or similar control channels) or semi-static, where test allocation information is indicated in the SIB and / or other broadcast messages.

[0152] WTRU can receive standard CSI-RS and / or apply some form of preprocessing, such as Figure 6 As shown in the diagram. Preprocessing may include extracting the regular CSI-RS received on the allocated resources. The WTRU may apply an OOD detection test to the extracted CSI-RS.

[0153] WTRU can receive L-RS and / or apply some kind of preprocessing, such as Figure 6 As shown in the diagram. Preprocessing may include extracting the L-RS received on the allocated resource. Preprocessing may include dividing by a known L-RS symbol or multiplying by the conjugate of a known L-RS symbol. Preprocessing may include a 2D FIR filtering option for the RE carrying the L-RS. The WTRU may apply an OOD detection test to the extracted L-RS.

[0154] WTRU can use L-RS and / or conventional CSI-RS for a channel estimator that uses high-quality and / or high-volume input data to generate high-quality labels (e.g., CSI labels) for an AI / ML CSI estimator, such as... Figure 6 As shown in the diagram, regular CSI-RS and / or CSI labels can be saved in the replay buffer for continued training.

[0155] WTRU can use regular CSI-RS as input to a trainable channel estimator (AI / ML-based channel estimator). WTRU can generate the output CSI, such as... Figure 6 As shown in the diagram, regular CSI-RS can be stored in the replay buffer.

[0156] WTRU can calculate the error between the CSI labels and / or the output CSI. If the error is below a threshold, WTRU can report the result to gNB and / or continue using the same AI / ML model. If the error is above the threshold, WTRU can report the result to gNB and / or request online retraining of the AI / ML model.

[0157] Figure 7An example AI / ML CSI estimation module 700 is described from the perspective of WTRU. The AI / ML channel estimator 704 receives CSI-RS 708 as input. After performing the operations described herein, the AI / ML channel estimator 704 generates an output as an estimated channel 712.

Claims

1. A method implemented by a wireless transmit / receive unit (WTRU), the method comprising: Receive pseudo-random data (PRD); Receive one or more Channel State Information Reference Signals (CSI-RS); Receive one or more learning reference signals (L-RS); The CSI measurement results are determined based on the received CSI-RS; The L-RS measurement results are determined based on the received L-RS. Generate one or more CSI tags based on PRD, CSI-RS measurement results, and L-RS measurement results; Use an artificial intelligence / machine learning (AI / ML) model and generate one or more output CSIs based on the PRD and the CSI-RS measurement results; as well as The AI / ML model is trained using the one or more CSI labels and the one or more output CSIs.

2. The method according to claim 1, further comprising: The change in the loss function of the AI / ML model is determined based on the one or more CSI labels and the one or more output CSIs; as well as The completion of training is determined by comparing the value of the loss function with a threshold.

3. The method of claim 2, wherein the loss function of the AI / ML model determines a measure of training accuracy during the training of the AI / ML model.

4. The method according to claim 1, further comprising: Determine the model drift metric; The performance of the AI / ML model used for CSI estimation is determined based on model drift by applying statistical tests to the CSI-RS measurement results. as well as The AI / ML model is retrained based on the model drift metric exceeding a drift threshold.

5. The method according to claim 4, further comprising: In response to an indication that the model drift metric exceeds the drift threshold, a report is sent to the network, wherein the report includes details of the model drift or recommendations for adjusting the AI / ML model.

6. The method according to claim 1, further comprising: A report is sent to the network based on the CSI-RS measurement results and the request to be configured with one or more training assignments; Receive from the network an indication as to whether the request to be configured with one or more training assignments is permitted.

7. The method according to claim 6, wherein, The instructions also include seed information for generating the PRD.

8. The method according to claim 6, further comprising: Receive PRD training allocation configuration, wherein the PRD training allocation indicates a first type of PRD training allocation, a second type of PRD training allocation, or a third type of PRD training allocation; The first type of PRD training assignment configures the WTRU to use PRD and CSI-RS measurements for a first set of resource blocks carrying PRD and CSI-RS. The second type of PRD training allocation configures the WTRU to use PRD and CSI-RS measurements for a second set of resource blocks containing data, including PRD, CSI-RS, and resource elements without PRD, wherein the PRD and the data-including resource elements without PRD occupy the same resource blocks within the second set of resource blocks; and The third type of PRD training allocation configures the WTRU to use PRD and CSI-RS measurements for a third set of resource blocks containing data, carrying PRD, CSI-RS, and not carrying PRD, wherein the PRD and the resource elements containing data without PRD occupy different but overlapping resource blocks within the third set of resource blocks.

9. The method of claim 8, wherein the allocation of PRD resources of the first, second and third types is signaled in the downlink control information (DCI), semi-statically via system information blocks or via broadcast messages.

10. The method according to claim 1, further comprising: Send a first message, the first message including an online training request, wherein the online training request indicates the number of resource blocks carrying a PRD for training, channel statistical measurement results, a channel diversification request, and the priority of the request; and Receive a second message, the second message including an online training response, wherein the online training response indicates the location of the resource block of the PRD, the allocated online training duration, the allocated online training start time, and the allocated online training L-RS.

11. A wireless transmit / receive unit (WTRU) comprising a processor and a memory, the processor and memory being configured to: Receive pseudo-random data (PRD); Receive one or more Channel State Information Reference Signals (CSI-RS); Receive one or more learning reference signals (L-RS); The CSI measurement results are determined based on the received CSI-RS; The L-RS measurement results are determined based on the received L-RS. Generate one or more CSI tags based on PRD, CSI-RS measurement results, and L-RS measurement results; Use an artificial intelligence / machine learning (AI / ML) model and generate one or more output CSIs based on the PRD and the CSI-RS measurement results; as well as The AI / ML model is trained using the one or more CSI labels and the one or more output CSIs.

12. The WTRU of claim 11, wherein the processor is further configured to: The change in the loss function of the AI / ML model is determined based on the one or more CSI labels and the one or more output CSIs; and The completion of training is determined by comparing the value of the loss function with a threshold.

13. The WTRU of claim 12, wherein the loss function of the AI / ML model determines a measure of training accuracy during the training of the AI / ML model.

14. The WTRU of claim 11, wherein the processor is further configured to: Determine the model drift metric; The performance of the AI / ML model used for CSI estimation was determined based on model drift by applying statistical tests to the CSI-RS measurement results; and The AI / ML model is retrained based on the model drift metric exceeding a drift threshold.

15. The WTRU of claim 14, wherein the processor is further configured to: In response to an indication that the model drift metric exceeds the drift threshold, a report is sent to the network, wherein the report includes details of the model drift or recommendations for adjusting the AI / ML model.

16. The WTRU of claim 1, wherein the processor is further configured to: A report is sent to the network based on the CSI-RS measurement results and the request to be configured with one or more training assignments; Receive from the network an indication as to whether the request to be configured with one or more training assignments is permitted.

17. The WTRU according to claim 16, wherein, The instructions also include seed information for generating the PRD.

18. The WTRU of claim 16, wherein the processor is further configured to: Receive PRD training allocation configuration, wherein the PRD training allocation indicates a first type of PRD training allocation, a second type of PRD training allocation, or a third type of PRD training allocation; The first type of PRD training assignment configures the WTRU to use PRD and CSI-RS measurements for a first set of resource blocks carrying PRD and CSI-RS. The second type of PRD training allocation configures the WTRU to use PRD and CSI-RS measurements for a second set of resource blocks containing data, including PRD, CSI-RS, and resource elements without PRD, wherein the PRD and the data-including resource elements without PRD occupy the same resource blocks within the second set of resource blocks; and The third type of PRD training assignment configures the WTRU to use PRD and CSI-RS measurements for a third set of resource blocks that include data, both carrying PRD and CSI-RS and without PRD. The PRD and the resource element including data without the PRD occupy different but overlapping resource blocks within the third resource block set.

19. The WTRU of claim 18, wherein the allocation of PRD resources of the first, second and third types is signaled in the downlink control information (DCI), semi-statically via system information blocks or via broadcast messages.

20. The WTRU of claim 11, wherein the processor is further configured to: Send a first message, the first message including an online training request, wherein, The online training request indicates the number of resource blocks carrying the PRD for training, channel statistics measurement results, channel diversification request, and request priority; as well as Receive a second message, the second message including an online training response, wherein the online training response indicates the location of the resource block of the PRD, the allocated online training duration, the allocated online training start time, and the allocated online training L-RS.

Images